aesara
aesara copied to clipboard
Aesara is a Python library for defining, optimizing, and efficiently evaluating mathematical expressions involving multi-dimensional arrays.
This file includes an alternative `Ger` `Op` that goes through scipy when in Python mode. This seems unnecessary, We have many more such routines implemented in `blas.py`, and it seems...
```python import aesara import aesara.tensor as at import numpy as np rng = aesara.shared(np.random.default_rng()) x = at.random.normal(rng=rng) f = aesara.function([], x, updates={rng: x.owner.outputs[0]}, mode="JAX") assert f() != f() # Fails...
The `environment-arm.yml` file is a copy of `environment.yml` with Intel MKL dependencies replaced with OpenBLAS. We may want to update the documentation on installing `aesara` on Mac OS or complete...
```python import aesara import aesara.tensor as at import numpy as np k = at.iscalar("k") A = at.vector("A") result, _ = aesara.scan(fn=lambda prior_result, A: prior_result * A, outputs_info=at.ones_like(A), non_sequences=A, n_steps=k) final_result...
```python import aesara import aesara.tensor as at x, y = at.scalars("x", "y") f = aesara.function([x, y], [x + y, x + y, y + x, y + x]) aesara.dprint(f) ```...
Importing Aesara yields the following error due to the `mkl` package. Recreating the conda environment in [`environment.yml`](https://github.com/aesara-devs/aesara/blob/main/environment.yml) minus the [`mkl` dependencies](https://github.com/aesara-devs/aesara/blob/main/environment.yml#L20-L22) allows Aesara to be imported. Given that `mkl` is...
Currently, we are not running CI with Mac OS so verifying continued correct behavior on OS X is more difficult. Github Actions provides access to Mac OS Intel machines so...
If people are interested, we could add string types to Theano. From some experimentation a while back, it seems like these types could be a thin wrapper around the built-in...
There are a few places in the documentation with "exercises" and "solutions" (e.g. [here](https://aesara.readthedocs.io/en/latest/tutorial/loop.html)). We don't need that form of presentation, so let's remove it.
Apparently, use of updates for random sampling within a `Scan` can be quite costly. The following illustrates: ```python import aesara import aesara.tensor as at from aesara.ifelse import ifelse import numpy...